# degrees of freedom

I’m having the hardest time remembering which dof to use for the various tests (e.g., n-2 vs. n-k-1, etc.) Any tips?

eek! normally isn’t it almost always n - (k+1) so the n - 2 just happens when you have 1 independent variable as for the n - k - 1… a lightbulb faintly is going off… F test… k and n - k - 1 df? i hope quant is not hard on this test or i am f’d.

F test = MSR/MSE MSR = RSS/k MSE = SSE/(n-(k+1) I hope i got the acroynms right.

banni, n-(k-1)=n-(k+1) just thinking various t-tests, f-tests, DW, chi-square, etc

for the serial correlation of the autoregression, its n-2 and for the regression of 1 independent variable its n-2. All the others I think are n-k-1. Thats how I’m remember them. forgot to mention the tests, all the tests (heterostastdcy, serial correlation, and multicolineary of a multiple regression are all based on k degrees of freedom)

In my personal notes I have for F test: Df numerator= k Df denom = n-1. Is denominator n-k-1? This is the kind of s&!t I don’t need 8 days out. Oh well, thanks for the correction if I am wrong.

F test --------- Num = k Den = n-k-1

bannisja Wrote: ------------------------------------------------------- > eek! > normally isn’t it almost always n - (k+1) > so the n - 2 just happens when you have 1 > independent variable n-2=n-(1+1) for one independent variable. Just so you don’t have to stress to try to remember…

Thanks, all. Just went to the Schweser notes, and you are right. (Red pen to my notes)

looking @ quicksheet: t-test for r: n-2 t-test for b in multiple regression: n-k-1 f-test: n-k-1 not sure about the others. i always get these messed up and i know cfa is going to have an answer where one df makes the difference btw. a right and wrong answer

D/W - n and k. Breush Pagan : just k.

D/W - n and k. Breusch Pagan : just k.

good thread… i was thinking of starting one. other good threads would be model assumptions and intercepts on different market models.

Dsylexic Wrote: ------------------------------------------------------- > D/W - n and k. Breush Pagan : just k. Yep…was just about to post this. Make sure!